This article analyzes reported data on length of stay of discharged patients from the 1977 National Nursing Home Survey. Assuming that patients admitted to nursing homes are one of two types, short-stayers and long-stayers, the statistically best-fitting proportions and expected lengths of stay for the two types are derived. The results are applied to statistics on characteristics of resident and discharged patients to find admission characteristics that differentiate the two groups. Because long-stayers and short-stayers have quite different characteristics, nursing home statistics will be very different if one bases them on discharges or a cross-section of residents. For example, we estimate that short-stay patients represent 61 per cent of discharges but only 9 per cent of residents. A further application shows that, including deaths in hospitals, almost half of discharged patients die, rather than the usually cited figure of 25 per cent.